Background Internet addiction (IA) is prevalent among adolescents but it is potentially revertible. Only three Taiwan adolescent studies reported IA remission and a few related factors. We investigated incidence and predictors of remission among Hong Kong Chinese secondary school students with a 12-month longitudinal study. Methods IA was defined as Chen Internet Addiction Scale (CIAS) score > 63. Validated measures were used to assess students' psychosocial wellbeing at baseline and follow-up. Results Of 1545 students with IA at baseline, 1296 (83.9%) provided matched baseline/12-month follow-up data; their data were analyzed. Incidence of remission (CIAS ≤ 63 at follow-up) was 59.29/100 person-years. Significant predictors included: 1) baseline CIAS score (ORa = .95), 2) baseline health belief model (HBM) constructs [perception of having severe IA (ORa = .34), perceived susceptibility to IA (ORa = 0.82), perceived barrier (ORa = 0.95), cue to action from parents (ORa = 0.82), and self-efficacy for reducing Internet use (ORa = 1.13)], and 3) baseline psychosocial health measures [self-esteem (ORa = 1.03), severe depression (ORa = 0.72) and social anxiety (ORa = 0.96)] and their changes over time [depression (ORa = .95), anxiety (ORa = .94), loneliness (ORa = .93), self-esteem (ORa = 1.07), positive affect (ORa = 1.10) and family support (ORa = 1.03)]. Two-thirds (64.3%) of the remission group presented reduced CIAS score > 1.5 SD, and recorded larger improvements in psychosocial status over time than the non-remission group. Conclusion Without noticeable interventions, incidence of remission was high and related to improvements in psychosocial health. Most of the HBM constructs, and baseline/changes in psychosocial measures predicted remission. Interventions to increase remission should modify these factors.